8206074: nsk/jdi/EventRequestManager/createStepRequest/crstepreq001/TestDescription.java is timing out

[Daniil Titov]

Hi Chris,

> and would not delete the single step request until the event has been received and the  thread is suspended.

In this test there is no guarantee that this single step event is ever posted.  In the most of the cases by the time the step request
is created the debuggee is already waiting on line 42 for blocking IO somewhere inside pipe.readlln(). And since the test sends the command to the pipe 
only after it deletes the step request the debuggee continues without posting the step event. 


  34	public class crstepreq001t {
    35	    public static void main(String args[]) {
    36	        ArgumentHandler argHandler = new ArgumentHandler(args);
    37	        IOPipe pipe = argHandler.createDebugeeIOPipe();
    38	        Thread  thr = Thread.currentThread();
    39	
    40	        thr.setName(crstepreq001.DEBUGGEE_THRD);
    41	        pipe.println(crstepreq001.COMMAND_READY);
    42	        String cmd = pipe.readln();
    43	        if (!cmd.equals(crstepreq001.COMMAND_QUIT)) {
    44	            System.err.println("TEST BUG: unknown debugger command: "
    45	                + cmd);
    46	            System.exit(crstepreq001.JCK_STATUS_BASE +
    47	                crstepreq001.FAILED);
    48	        }
    49	        System.exit(crstepreq001.JCK_STATUS_BASE +
    50	            crstepreq001.PASSED);
    51	    }
   


Best regards,
--Daniil

On 6/5/19, 7:44 PM, "Chris Plummer" <chris.plummer at oracle.com> wrote:

    Hi Daniil,
    
    I understand the general issues that the debugger might delete a request 
    after the event has already been generated but before the debugger has 
    received it. Alan and I recently had a discussion about this and 
    concluded that the debugger must be ready for this, otherwise the debug 
    agent will have suspended threads and the debugger might never do 
    anything to resume them (in fact I suspect I'm seeing this in a certain 
    debugger). However, for the case of this test I would think it would be 
    taking fine grain controlled over the requests and events, and would not 
    delete the single step request until the event has been received and the 
    thread is suspended. It can then delete the request and resume the 
    thread (or all threads if using SUSPEND_ALL). Is there a reason the test 
    is not already doing this?
    
    thanks,
    
    Chris
    
    On 6/5/19 7:36 PM, Daniil Titov wrote:
    > Please review a change that fixes the intermittent failure of the test.
    >
    > The problem here that there is a chance that a single step event had been posted after the step request was created and before it was deleted.
    >
    > The fix solves this by ensuring that vm.resume() is called at the end of the test.
    >
    > Webrev: http://cr.openjdk.java.net/~dtitov/8206074/webrev.01
    > Bug: https://bugs.openjdk.java.net/browse/JDK-8206074
    >
    > Thanks!
    > --Daniil
    >
    >
    >